Stratégie évolutivement stable

En théorie des jeux, en psychologie comportementale et en psychologie évolutionniste, une stratégie évolutivement stable ou SES (en anglais, evolutionarily stable strategy ou ESS) est un cas particulier d'équilibre de Nash tel que, dans une grande population de joueurs se rencontrant aléatoirement, plusieurs stratégies peuvent coexister chacune possédant une fréquence d'équilibre propre. Développé originellement en 1973 par John Maynard Smith et George R. Price pour expliquer l'évolution des comportements sociaux, cette théorie est maintenant largement utilisée en écologie comportementale, en sociobiologie, en économie, en psychologie évolutionniste, en anthropologie biosociale, en science politique et en philosophie. Cette théorie fut exposée au grand public en 1976 par Richard Dawkins dans son best-seller Le Gène égoïste et elle fut récompensée en 1999 par un prix Crafoord. Historique La notion de stratégie évolutionnairement stable (SES) a été introduite par John Mayna

A heterogeneous evolutionary stable population under assortative matching: Exploring the diversity of preferences

Charles Chadi Ayoubi, Boris Thurm

Recent studies have shown that a population acting not only upon self-interest but also exhibiting some morality preference has an evolutionary advantage. Specifically, in the setting of a symmetric fitness game, a resident population is evolutionary stable against all types of mutants if it has the utility function of Homo-moralis, with a morality equal to the assortativity. In this paper, we extend the scope of analysis allowing for the presence of a diversity of preferences in the population. Establishing a Payoff Equality condition, we prove the possibility of co-existence of two residents in the population. We then introduce a tripartite assortment function and study the conditions for the evolutionary stability of this diverse population. In the case of a constant assortment function, we show the existence of an evolutionary stable and heterogeneous resident population.

2017

The diversity of moral preferences: Evolutionary foundations and some implications in environmental economics

Boris Thurm

Empirical evidence suggests that there exists substantial heterogeneity in individuals' social preferences. However, there is little theoretical basis supporting this observation and economic models often assume that all individuals are identical. Hence, the aim of this thesis is to provide theoretical foundations for the observed heterogeneity of social preferences and to derive its implications for environmental policy. We first extend the framework of evolutionary game theory introducing the concepts of evolutionarily stable population and an assortment matrix to study the evolution of preferences in assortatively matched interactions between heterogeneous individuals. We show that there exists a heterogeneous evolutionarily stable population composed of both fully-selfish and fully-moral individuals for some but not all games and assortment structures. Therefore, the preferences that are favored by evolution depend on the socio-economic environment. In particular, our analysis highlights the key role played by the assortment structure in the existence and the robustness of heterogeneous evolutionarily stable populations. We then design a model with heterogeneous moral individuals involved in a social dilemma. Our framework sheds light on many empirical findings explaining why some individuals are willing to voluntarily engage in costly pro-environmental actions even though the impact of their efforts on environmental externalities is negligible. Investigating how individuals' beliefs can alter their behaviors and hinder cooperation, we demonstrate why financial incentives can fail to foster pro-environmental behaviors in some cases while non-financial incentives such as nudges and educational campaigns could be successful. Consequently, better accounting for the social motives behind individuals' decisions in economic models could help policy makers design more effective policies.

EPFL2019

Learning from demonstration using products of experts: Applications to manipulation and task prioritization

Sylvain Calinon, Emmanuel Pignat

Probability distributions are key components of many learning from demonstration (LfD) approaches, with the spaces chosen to represent tasks playing a central role. Although the robot configuration is defined by its joint angles, end-effector poses are often best explained within several task spaces. In many approaches, distributions within relevant task spaces are learned independently and only combined at the control level. This simplification implies several problems that are addressed in this work. We show that the fusion of models in different task spaces can be expressed as products of experts (PoE), where the probabilities of the models are multiplied and renormalized so that it becomes a proper distribution of joint angles. Multiple experiments are presented to show that learning the different models jointly in the PoE framework significantly improves the quality of the final model. The proposed approach particularly stands out when the robot has to learn hierarchical objectives that arise when a task requires the prioritization of several sub-tasks (e.g. in a humanoid robot, keeping balance has a higher priority than reaching for an object). Since training the model jointly usually relies on contrastive divergence, which requires costly approximations that can affect performance, we propose an alternative strategy using variational inference and mixture model approximations. In particular, we show that the proposed approach can be extended to PoE with a nullspace structure (PoENS), where the model is able to recover secondary tasks that are masked by the resolution of tasks of higher-importance.

SAGE PUBLICATIONS LTD2021

The diversity of moral preferences: Evolutionary foundations and some implications in environmental economics

Boris Thurm

EPFL2019